Operations on masks

Given a Spectrum, Image or Cube object, O, the pixel values and variances of this object can be accessed via the O.data and O.var arrays. These are masked arrays, which are arrays that can have missing or invalid entries. Masked arrays, which are provided by the numpy.ma module, provide a nearly work-alike replacement for numpy arrays. The O.data and O.var arrays share a single array of boolean masking elements. This is a simple boolean array, and it can be accessed directly via the O.mask property. The elements of this array can be modified implicitly or explicitly via operations on the data, variance or mask arrays. For example:

In [1]: import numpy as np

In [2]: import matplotlib.pyplot as plt

In [3]: from mpdaf.obj import Image

In [4]: ima = Image('obj/a370II.fits')

In [5]: ima.data[1000:1500, 800:1800] = np.ma.masked

# is equivalent to
In [6]: ima.mask[1000:1500, 800:1800] = True

In [7]: ima.plot(vmin=1950, vmax=2400, colorbar='v')
Out[7]: <matplotlib.image.AxesImage at 0x7f58199a87d0>

The comparison operators can also be used to mask all data that are not less than or equal (<=) to a value, are not less than (<) a value, are not greater than or equal (>=) to a value, or that are not greater than (>) a given value. In the following example, the < operator is used to select data that have values over 2000, and mask everything below this threshold. In the plot of the resulting image, the masked values below 2000 are drawn in white. These values would otherwise be dark blue if they weren’t masked.

In [8]: ima.unmask()

In [9]: ima2 = ima > 2000

In [10]: plt.figure()
Out[10]: <matplotlib.figure.Figure at 0x7f581916f4d0>

In [11]: ima2.plot(vmin=1950, vmax=2400, colorbar='v')
Out[11]: <matplotlib.image.AxesImage at 0x7f5811f42750>

Note the use of the unmask method to clear the current mask.

It is also possible to mask pixels that correspond to a selection, by using the mask_selection method, or to mask pixels whose variance exceeds a given threshold value, by using the mask_variance method:

In [12]: ima.unmask()

In [13]: ksel = np.where(ima.data < 2000)

In [14]: ima.mask_selection(ksel)

In [15]: plt.figure()
Out[15]: <matplotlib.figure.Figure at 0x7f5819306e10>

In [16]: ima.plot(vmin=1950, vmax=2400, colorbar='v')
Out[16]: <matplotlib.image.AxesImage at 0x7f58198a0c50>

On Image or Cube objects, there are additional methods to mask inside or outside:

For example:

In [17]: ima.unmask()

In [18]: ima.mask_region(center=[800.,600.], radius=500., unit_center=None, unit_radius=None, inside=False)

In [19]: plt.figure()
Out[19]: <matplotlib.figure.Figure at 0x7f5819388fd0>

In [20]: ima.plot(vmin=1950, vmax=2400, colorbar='v')
Out[20]: <matplotlib.image.AxesImage at 0x7f5812007690>